We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:

Download:

Current browse context:

cs.LG

Change to browse by:

References & Citations

DBLP - CS Bibliography

Bookmark

(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo ScienceWISE logo

Computer Science > Machine Learning

Title: Variational Bayesian Inference for Hidden Markov Models With Multivariate Gaussian Output Distributions

Abstract: Hidden Markov Models (HMM) have been used for several years in many time series analysis or pattern recognitions tasks. HMM are often trained by means of the Baum-Welch algorithm which can be seen as a special variant of an expectation maximization (EM) algorithm. Second-order training techniques such as Variational Bayesian Inference (VI) for probabilistic models regard the parameters of the probabilistic models as random variables and define distributions over these distribution parameters, hence the name of this technique. VI can also bee regarded as a special case of an EM algorithm. In this article, we bring both together and train HMM with multivariate Gaussian output distributions with VI. The article defines the new training technique for HMM. An evaluation based on some case studies and a comparison to related approaches is part of our ongoing work.
Comments: Preliminary version. Contains all necessary equations for implementation. Ongoing research
Subjects: Machine Learning (cs.LG); Machine Learning (stat.ML)
Cite as: arXiv:1605.08618 [cs.LG]
  (or arXiv:1605.08618v1 [cs.LG] for this version)

Submission history

From: Christian Gruhl [view email]
[v1] Fri, 27 May 2016 13:00:31 GMT (2967kb,D)

Link back to: arXiv, form interface, contact.